【正文】 on of the model. In this case no such data was available. Dividing the existing data set into two portions to estimate the model on one portion and use the other for validation was not practical, given the limited sample size in some of the submodels. For example, the concrete pavement submodel has a total of only 212 observations, and estimating the submodel on the highly variable data on fewer observations would reduce the accuracy of the estimates. Thus, the performance of the model was assessed by observing how well it reproduced observed construction costs. Using the same data as that on which the model was calibrated, the estimated and observed LHCI values for the period 1984–1997 are shown in Fig. 3. The 95% confidence limit of the observed LHCI is also shown in the figure to illustrate that the estimated LHCI values are, for the most part, contained within the 95% confidence limit of the observed LHCI values. The chisquared test of the similarity of the estimated and observed LHCI values indicates that a significant difference could not be observed at the 99% level of significance. Investigating the behavior of the construction cost index in Fig. 3 reveals interesting reasons behind the observed behavior. Reviewing the data and observing its impact on the forecasts through the model allows an analyst to determine the primary causes of change in construction costs during certain periods in the past. For example, the main cause of the decrease in construction costs observed in the period 1984–1986 can be traced back to a decline in labor and petroleum costs during that period. The rapid increase in construction costs from 1995 to 1996 was primarily due to a bination of rising petroleum costs and an increased proportion of smaller contracts. The drop in construction costs observed immediately following this event (., in 1997) was mainly the consequence of an increase in the average size of projects from those let in 1996, very few projects being let in the fourth quarter, and a decrease in the average duration of projects. Conclusions This study has shown that the literature indicates that a prehensive set of factors contributes to the cost of highway construction. In this study, the most influential factors were found to be the cost of the material, labor, and equipment used in constructing the facility. However, characteristics of individual contracts and the contracting environment in which contracts are let also affect construction costs. In particular, contract size, duration, location, and the quarter in which the contract is let were found to have a significant impact on contract cost. Bid volume, bid volume variance, number of plan changes, and changes in construction practice, standards, or specifications also make a significant impact on contract costs. Other factors are expected to have an impact on construction costs but were not included in this analysis because no data on their values were available. The model developed in this study reproduces past overall construction costs reasonably accurately at the aggregate level. Predicted overall construction costs are not significantly different from observed costs at the 99% level of significance. This accuracy is largely the result of the aggregate level at which construction costs are measured in this study。 本 科 生 畢 業(yè) 設(shè) 計（論文） 外文翻譯 題目Ⅰ： Estimating Future Highway Construction Costs Estimating Future Highway Construction Costs C. G. Wilmot, ,1 and G. Cheng, Abstract: The objective of this research was to develop a model that estimates future highway construction costs in Louisiana. The model describes overall highway construction cost in terms of a highway construction cost index. The index is a posite measure of the cost of construction labor, materials, and equipment。 at the individual contract level, the submodels capture only between 42 and 72% of the variation in the data. It is suspected that much of this variation is due to unobserved, essentially subjective factors that influence the bid prices in individual contracts. However, some of these idiosyncratic variations at the individual contract level average out in the aggregation process. This model can be used by highway officials in Louisiana to test alternative contract management strategies. Increasing contract sizes, reducing the duration of contracts, reducing bid volume and bid volume variance, reducing the number of plan changes, and reducing the proportion of contracts let in the fourth quarter all serve to reduce overall construction costs. Highway officials can assess the impact of strategies they believe are achievable by applying the model. Most importantly, though, the model can assist in estimating future construction costs and providing the means to produce more reliable construction programs. Reference Associate Professor, Louisiana Transportation Research Center and Dept. of Civil and Environmental Engineering, Louisiana State Univ., Baton